Ten prompts for generating your own exam-level questions. Swap the placeholders for the areas you find hardest.
How to use these. Swap only the words in double braces, such as {{DOMAIN}}. Keep the structure and rules intact, because the structure is what makes the output exam-like. Paste the result into NotebookLM, Gemini or another AI tool. Treat every legal reference an AI gives you as a claim to check, and verify it against the source text.
Placeholders you can swap
Placeholder
What to put there
Example
{{DOMAIN}}
The exam domain you are weakest in
Domain III: AI development
{{TOPIC}}
The specific topic inside that domain
validation of models across subgroups
{{NUMBER}}
How many items you want
10
{{DIFFICULTY}}
Target hardness
Very hard, exam level
{{JURISDICTION}}
The legal frame to focus on
EU AI Act and GDPR
{{MY_BACKGROUND}}
Your starting point, so the questions match what you already know
Five years in privacy, new to model testing
{{QUESTION_STYLE}}
The format you want
Scenario, four options, one best answer
{{SOURCE_TEXT}}
Text you own and want questions drawn from (optional)
Paste your own notes
{{TRAP_TYPE}}
The trap pattern to practise
Role misallocation (provider vs deployer)
{{SELECT_COUNT}}
How many answers a multi-select item needs
3
{{SUBJECT}}
The area for NOT-style items
monitoring after deployment
{{QUESTION}}
A practice question you got wrong, pasted in full
(paste)
{{MY_ANSWER}}
The answer you chose, and why if you know
(paste)
{{CORRECT}}
The correct answer, if you have it
(paste, or leave blank)
{{SCENARIO}}
The scenario for a short-answer question
(paste)
{{SCORES_BY_DOMAIN}}
Your latest practice scores by domain
Domain I 70%, II 55%, III 60%, IV 48%
{{DAYS_LEFT}}
Days until your exam
60
{{HOURS_PER_WEEK}}
Hours you can study each week
8
{{ARTICLE}}
The article or clause to drill
Article 6(3)
{{REGULATION}}
The regulation or standard it belongs to
Regulation (EU) 2024/1689
The prompts
1. Exam-level question writer for your weak areas
The core prompt. It produces hard, scenario-based questions with defensible distractors, and it checks its own work before answering.
You are an experienced item writer for a professional AI governance certification. You write scenario-based questions that test application and analysis, not recall.
CONTEXT
- Target domain: {{DOMAIN}}
- Topic: {{TOPIC}}
- My background: {{MY_BACKGROUND}}
- Jurisdiction focus: {{JURISDICTION}}
- Difficulty: {{DIFFICULTY}}
- Number of questions: {{NUMBER}}
- Format: {{QUESTION_STYLE}}
RULES FOR EACH QUESTION
1. Write a scenario of 80 to 140 words that contains one concrete decision.
2. Ask one question. The final sentence must name the exact role, jurisdiction or lifecycle stage the question is about.
3. Give four options with exactly one best answer. Each distractor must be defensible in part: a true statement that answers a different question, or a correct action taken by the wrong role or at the wrong stage.
4. Do not use "all of the above", "none of the above", or options that are noticeably longer or shorter than the others.
5. Do not repeat the wording of the correct answer in the stem.
6. Where an item depends on a legal article or framework clause, write [VERIFY: source] unless you are certain of the reference.
OUTPUT FOR EACH QUESTION
- Stem
- Options A to D
- Correct answer
- Rationale (3 to 5 sentences)
- Distractor analysis: one sentence per wrong option, saying why it is tempting and why it fails
- Source to check
- Confidence from 1 to 5, with one line on why
BEFORE YOU OUTPUT
Re-read every answer against its rationale. Rewrite any item where a second option could also be defended. Then list the items you are least sure of.
If I have provided {{SOURCE_TEXT}}, use only that text for facts, and write NOT IN SOURCE where it does not cover something.
Before you start, state any assumptions you are making.
2. Trap-style question builder
Trains you to spot the specific distractor patterns exam writers tend to use.
You are writing distractors for a professional certification exam. Build {{NUMBER}} questions around this trap pattern: {{TRAP_TYPE}}.
For each question:
(a) State the correct concept in one sentence.
(b) Write a scenario in which a candidate who has memorised a slogan would choose the wrong answer.
(c) Write four options in which the trap answer is the most attractive.
(d) Name the exact word or phrase in the stem that decides the answer.
(e) Mark any legal or standards claim as [VERIFY].
Do not reuse wording from any official exam material. After the set, write a two-sentence rule I can apply to recognise this trap in future questions.
3. Multi-select and NOT-style questions
Builds the two formats that catch candidates who skim the question stem.
Write {{NUMBER}} multi-select questions. Each gives five options, and the candidate must select exactly {{SELECT_COUNT}}. Then write {{NUMBER}} NOT-style questions about {{SUBJECT}}, where the correct answer is the one that does not belong.
Rules:
- State the rule that defines the correct set in the stem itself, so the candidate can apply it.
- For NOT-style items, make every other option satisfy the rule, so only one fails.
- Avoid trick wording such as double negatives.
- After each item, give the answer and a one-line explanation.
Before you output, check that each multi-select item has exactly {{SELECT_COUNT}} correct options.
4. Socratic tutor after a wrong answer
Turns a mistake into a lesson by finding the assumption behind it, rather than just giving the right answer.
You are a patient tutor. I answered a practice question incorrectly.
QUESTION: {{QUESTION}}
MY ANSWER: {{MY_ANSWER}}
CORRECT ANSWER (if known): {{CORRECT}}
Work in steps and do not skip ahead.
Step 1: Do not reveal the correct answer. Ask me two questions that expose my reasoning.
Step 2: After I reply, name the exact assumption that led to my choice.
Step 3: Explain the concept in plain English, with one analogy from everyday work.
Step 4: Write one new question at the same difficulty, on the same concept, with a different scenario.
Stop after Step 1 and wait for my reply.
5. Distractor autopsy
Teaches you to read options the way the exam writer intended, which is the fastest way to improve on scenario questions.
Here is a practice question:
{{QUESTION}}
Options are included above. Do the following:
1. Rank the three wrong options from most to least tempting.
2. For each wrong option, describe the reasoning path that makes it attractive, and the specific fact that eliminates it.
3. Identify the two words in the stem that were the strongest clue to the answer.
4. Write one reusable rule for similar questions.
Do not change the question. If any option could be defended, say so and explain why.
6. Source-grounded quiz from your own notes
Keeps every question tied to a text you own, so you can check each answer against it.
Use only the source text provided, titled {{SOURCE_NAME}}. Do not add facts from elsewhere.
Write {{NUMBER}} questions at {{DIFFICULTY}} level. For each question:
- State the section heading where the answer is found.
- Give the correct answer and the sentence that supports it, paraphrased.
- If the topic is not covered by the source, do not write a question about it. Write NOT IN SOURCE instead.
After the quiz, list the topics an exam on this subject would normally test that the source does not cover.
SOURCE TEXT:
{{SOURCE_TEXT}}
7. Scenario short-answer marker
Gives you a strict, criteria-based mark on written answers, with hints rather than model answers.
Act as a strict but fair examiner.
SCENARIO: {{SCENARIO}}
MY ANSWER: {{MY_ANSWER}}
Score my answer from 1 to 5 on each criterion:
1. Identifies the correct role or party.
2. Names the relevant obligation or control.
3. Puts the steps in the right order.
4. States the trade-off or residual risk.
5. Uses the right terminology.
For each score, quote the phrase in my answer that earned or lost the mark. Then name the single most important missing point.
Do not rewrite my answer. Give me a hint, not a model answer.
8. Weak-area study planner
Spreads your study time across domains in proportion to both their exam weight and your weakness.
I have {{DAYS_LEFT}} days until my exam and {{HOURS_PER_WEEK}} hours a week to study.
My latest practice scores by domain: {{SCORES_BY_DOMAIN}}
Using the Body of Knowledge question ranges as weights (Domain I 16 to 20, Domain II 19 to 23, Domain III 21 to 25, Domain IV 21 to 25 questions), allocate my study time. Weight it toward my lowest scores, but give no domain less than 15% of my time.
Produce:
- A stated list of your assumptions, and anything you cannot confirm.
- A week-by-week plan.
- A 20-minute error-review block each day.
- A timed mini-test every seven days.
Keep the plan realistic for my weekly hours.
9. Regulation article drill
Tests conditions, exceptions and cross-references in a provision, not just its headline.
Take {{ARTICLE}} of {{REGULATION}}.
Write {{NUMBER}} questions that test its conditions, exceptions and cross-references rather than its headline. For each question:
- Point to the paragraph that answers it.
- Quote no more than 15 words from the text.
- If the answer depends on a definition elsewhere in the text, name that definition.
Flag any item where the text is ambiguous, or where a later amendment could change the answer, with [CHECK CURRENT TEXT].
10. Confidence calibration
Shows where you are confidently wrong, which is more dangerous on exam day than where you are simply unsure.
I will answer a set of {{NUMBER}} practice questions. For each one, I will give my answer and a confidence rating from 1 to 5 before you reveal the result.
After the set, produce a calibration report:
- My accuracy at each confidence level.
- Each question where I was confident (4 or 5) and wrong.
- The three concepts I should re-study first.
- A rule for when I should slow down and re-read the question.
Do not reveal the answers until I have answered all of them.
Independent study guide. Not affiliated with, endorsed by or sponsored by the IAPP, Dr. Kyle David or Dr. David Privacy. Not legal advice. About this site